Oddbean new post about login | logout "Using Side Eye, a machine learning assisted tool that Fu and his research team created, Fu can determine the gender of someone speaking in the room where a photo was taken –– and even the exact words they spoke." https://news.northeastern.edu/2023/09/25/audio-recovery-still-images-silent-videos/
This research blew my mind... basically they're using image stabilization technology in modern phones to see how camera lenses adjust themselves to sound waves (spoken words) and then extracting that information from photos, reconstructing the sound. 🤯 🤯 🤯 🤯 🤯 🤯 🤯
To save everyone from going and reading the paper to find what the catch is: it seems for this to work they had the phone sitting on the same rigid surface as the speaker producing the sound to be recognized. So don’t worry about people going through your past photos that you shot with a handheld camera. For now. 😄
@a6ff0b2b I was surprised to see you repost the Side Eye link – in particular, the remark about gender – without comment. This is an important clarification: https://infosec.exchange/@dymaxion/111147485741674480